Histogram Analysis of Small Solid Renal Masses: Differentiating Minimal Fat Angiomyolipoma From Renal Cell Carcinoma

Transcription

1 Genitourinary Imaging Original Research Chaudhry et al. Histogram Analysis of Small Solid Renal Masses Genitourinary Imaging Original Research Humaira S. Chaudhry 1,2 Matthew S. Davenport 1,3 Christopher M. Nieman 1,4 Lisa M. Ho 1 Amy M. Neville 1 Chaudhry HS, Davenport MS, Nieman CM, Ho LM, Neville AM Keywords: angiomyolipoma, clear cell renal cell cancer, CT histogram, histogram analysis, minimal fat angiomyolipoma, papillary renal cell cancer, renal cell carcinoma DOI: /AJR Received March 20, 2011; accepted after revision June 16, Department of Radiology, Division of Abdominal Imaging, Duke University Medical Center, Durham, NC. 2 Present address: Department of Radiology, Division of Abdominal Imaging, University of Medicine and Dentistry of New Jersey, New Jersey Medical School, University Hospital Suite C-318, 150 Bergen St, Newark, NJ Address correspondence to H. S. Chaudhry (hchaudry27@gmail.com). 3 Present address: Department of Radiology, University of Michigan Health System, University Hospital, Ann Arbor, MI. 4 Present address: Department of Radiology, Cleveland Clinic, Fairview Hospital, Fairview, OH. AJR 2012; 198: X/12/ American Roentgen Ray Society Histogram Analysis of Small Solid Renal Masses: Differentiating Minimal Fat Angiomyolipoma From Renal Cell Carcinoma OBJECTIVE. The objective of our study was to retrospectively determine whether minimal fat renal angiomyolipoma can be differentiated from clear cell or papillary renal cell carcinoma (RCC) in small renal masses using attenuation measurement histogram analysis on unenhanced CT. MATERIALS AND METHODS. Twenty minimal fat renal angiomyolipomas were compared with 22 clear cell RCCs and 23 papillary RCCs using an institutional database. All masses were histologically confirmed and all minimal fat renal angiomyolipomas lacked radiographic evidence of macroscopic fat. Using attenuation measurement histogram analysis, two blinded radiologists determined the percentage of negative pixels within each renal mass. The percentages of negative pixels below attenuation thresholds of 0, 5, 10, 15, 20, 25, and 30 HU were recorded. Sensitivity, specificity, positive predictive value, negative predictive value, and receiver operator characteristic curves for the diagnosis of minimal fat renal angiomyolipoma were generated for each threshold. The Student t test was used to compare radiologists and cohorts. Previously published attenuation and pixel-counting thresholds reported as having a specificity of near 100% for discriminating between minimal fat renal angiomyolipomas and RCCs were analyzed. RESULTS. The mean maximal transverse lesion diameter was 1.8 cm for minimal fat renal angiomyolipomas (SD, 0.5 cm; range, cm), 2.1 cm for clear cell RCCs (SD, 0.5 cm; range, cm), and 2.1 cm for papillary RCCs (SD, 0.7 cm; range, cm). No significant difference in the percentage of negative pixels was found between minimal fat renal angiomyolipomas and clear cell RCCs or between minimal fat renal angiomyolipomas and papillary RCCs at any of the selected attenuation thresholds for either radiologist (p = ). Radiologist 1 and radiologist 2 used significantly different region-of-interest sizes (p < 0.001), but neither radiologist could differentiate minimal fat renal angiomyolipoma from RCC. No previously published threshold allowed discrimination between minimal fat renal angiomyolipoma and RCC with 100% specificity. CONCLUSION. Attenuation measurement histogram analysis cannot reliably differentiate minimal fat renal angiomyolipoma from RCC. S olid renal masses are increasingly identified as incidental findings on cross-sectional imaging studies and can be either benign or malignant [1]. Angiomyolipoma is the most common benign solid renal tumor, with a prevalence of % in the general population [2]. Of the malignant neoplasms, clear cell renal cell carcinoma (RCC) is the most common, followed by the papillary subtype, which is also known as the chromophil subtype [3]. Most renal angiomyolipomas are diagnosed by the presence of macroscopic fat visible on cross-sectional imaging. CT and MRI are commonly used to make that assessment. An uncommon variant of angiomyolipoma is the minimal fat angiomyolipoma, which does not show macroscopic fat; thus, it can be difficult to differentiate from RCC on imaging [4]. This subtype has been noted in up to 5% of renal angiomyolipomas [5]. It is also known that small renal masses are, in general, more likely to be benign than large renal masses; for example, Frank et al. [6] reported that 22% of renal masses 2 3 cm (75/341) were benign, whereas only 5% of renal masses 6 7 cm (11/243) were benign. The ability to distinguish between benign and malignant masses by imaging criteria alone would be helpful. AJR:198, February

2 Chaudhry et al. In multiple previous studies, investigators have examined the use of imaging techniques to diagnose minimal fat renal angiomyolipomas with discordant results. The use of CT histogram analysis is advocated by Simpfendorfer et al. [7], Simpson and Patel [8], and Kim et al. [9] but is disputed by Catalano et al. [10]. Chemical-shift MRI has also been investigated as an imaging tool to diagnose minimal fat renal angiomyolipoma [11]; however, loss of signal intensity on opposed-phase imaging can also be seen with clear cell RCC, limiting the use of that modality [12]. The purpose of our study was to retrospectively determine whether minimal fat renal angiomyolipoma can be differentiated from clear cell or papillary RCC in small renal masses using attenuation measurement histogram analysis on unenhanced CT. Materials and Methods Institutional review board approval was obtained, and patient informed consent was waived for this retrospective analysis. Our study was compliant with HIPAA. Patient Selection The institutional pathology database was queried to identify all histologically confirmed renal angiomyolipomas obtained by biopsy or surgical resection from August 2002 through April Masses without digital unenhanced CT images before surgical intervention available on our institutional PACS were excluded. Two authors, a fellow and an abdominal imaging attending physician with 7 years experience at the time, reviewed the digital images and excluded masses with macroscopic fat (identified by visual assessment on CT datasets). This review resulted in 20 minimal fat renal angiomyolipomas in 19 patients (16 women and three men; mean age, 54 years; age range, years). The same authors also recorded the average attenuation value of each lesion by consensus and reviewed the electronic medical records of all patients to identify those with tuberous sclerosis. After establishing the minimal fat renal angiomyolipoma cohort, we identified similar cohorts of patients with clear cell RCCs and of patients with papillary RCCs who presented during the same time period. We searched the same institutional database for patients with histologically proven clear cell RCC or papillary RCC based on lesion size and patient demographics similar to those of our minimal fat renal angiomyolipoma cohort. This process was performed by one author who was blinded to the attenuation measurement histogram results. Lesions with evidence of concomitant metastatic disease were excluded from the cohort. This search resulted in 22 clear cell RCCs in 22 patients (10 women and 12 men; mean age, 57 years; age range, years) and 23 papillary RCCs in 23 patients (nine women and 14 men; mean age, 59 years; age range, years). After selecting cases for the clear cell RCC and papillary RCC cohorts, both authors recorded the average attenuation value for each lesion in the cohort by consensus. All RCCs were solid masses and an unenhanced CT study obtained before surgical intervention was available on our institutional PACS. If a patient underwent multiple CT examinations, we used only the most recent study. CT Examination Each mass was imaged with unenhanced CT performed on a variety of scanners at our institution. These scanners included a single-detector CT unit (HiSpeed CT/i, GE Healthcare; 140 kvp, ma), 4-MDCT unit (LightSpeed QX/i, GE Healthcare; 140 kvp, ma), 8-MDCT unit (LightSpeed Ultra, GE Healthcare; 140 kvp, 340 ma), 16-MDCT unit (LightSpeed 16, GE Healthcare; 140 kvp, ma), and 64-MDCT unit (LightSpeed VCT, GE Healthcare; 140 kvp, ma). The slice thickness for all examinations was 5 mm. Each scanner was calibrated daily using a standard water phantom. Six studies (minimal fat renal angiomyolipoma, n = 1; clear cell RCC, n = 3; papillary RCC, n = 2) were performed on single-detector CT units, 25 studies (minimal fat renal angiomyolipoma, n = 9; clear cell RCC, n = 5; papillary RCC, n = 11) on 4-MDCT units, one study (clear cell RCC, n = 1) on an 8-MDCT unit, 27 studies (minimal fat renal angiomyolipoma, n = 9; clear cell RCC, n = 11; papillary RCC, n = 7) on 16-MDCT units, and six studies (minimal fat renal angiomyolipoma, n = 1; clear cell RCC, n = 2; papillary RCC, n = 3) on 64- MDCT units. Histogram Analysis The same two authors reviewed each unenhanced CT study and identified by consensus which axial slice best showed each target solid renal mass. They also recorded the maximum transverse diameter of each lesion by consensus. A list of each of the masses with corresponding image numbers and diameters was generated and submitted to two board-certified radiologists for review. Both radiologists, an abdominal imaging fellow and an abdominal imaging attending physician with 12 years experience, were blinded to the histologic diagnosis. The two latter radiologists were directed to identify each renal mass using the provided image numbers. They then independently drew a region of interest (ROI) within each mass on any slice that they thought was ideal for analysis, similar to how they would evaluate the study in clinical practice. The image numbers helped the radiologists identify each mass, but the radiologists were not constrained to these image numbers during their analysis. All ROIs were drawn on a dedicated measurement system (Syngo Multimodality WorkPlace Series 8670, Siemens Healthcare). The software calculated the total number of pixels within each ROI and the ROI area (mm 2 ). Each radiologist recorded these data as well as the number of pixels with attenuation measurements equal to or below seven unique thresholds: 30, 25, 20, 15, 10, 5, and 0 HU. Statistical Analysis Mean mass size and mean attenuation, SD, and range were calculated for each cohort (minimal fat renal angiomyolipomas, clear cell RCCs, papillary RCCs). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for diagnosing minimal fat renal angiomyolipoma were calculated for multiple attenuation thresholds ( 30, 25, 20, 15, 10, 5, and 0 HU). Receiver operating characteristic (ROC) curves were derived for each attenuation and pixelcounting threshold combination. The Student t test was used to compare the attenuation measurement histogram results of radiologist 1 versus radiologist 2, papillary RCCs versus minimal fat renal angiomyolipomas, clear cell RCCs versus minimal fat renal angiomyolipomas, and papillary RCCs and clear cell RCCs versus minimal fat renal angiomyolipomas. The statistical analysis was completed utilizing Microsoft Excel. A p value of < 0.05 was chosen to indicate statistical significance. The number of masses that met previously published attenuation measurement histogram analysis criteria for the diagnosis of minimal fat renal angiomyolipoma [7] was recorded. These two criteria were 20 pixels or more with an attenuation of less than 20 HU, which was referred to in [7] as criterion 2, and 5 pixels or more with an attenuation of less than 30 HU, or criterion 3. Simpfendorfer et al. [7] reported that these criteria had 100% specificity and 100% PPV for the diagnosis of minimal fat renal angiomyolipoma. To remain consistent with their study, these criteria are also labeled in our study as criterion 2 and criterion 3. Results The mean maximal transverse diameter was 1.8 cm for minimal fat renal angiomyolipomas (SD, 0.5 cm; range, cm), 2.1 cm for clear cell RCCs (SD, 0.5 cm; range, cm), and 2.1 cm for papillary RCCs (SD, 0.7 cm; range, cm). One patient with one minimal fat renal angiomyolipoma 378 AJR:198, February 2012

3 Histogram Analysis of Small Solid Renal Masses in our study group had a history of tuberous sclerosis. Table 1 lists the patient demographics, scan characteristics, and ROI sizes used for each cohort. The unenhanced mean attenuation values of the masses for the cohorts of this study (minimal fat renal angiomyolipoma = 34 HU [range, HU]; clear cell RCC = 36 HU [range, HU]; papillary RCC = 38 HU [range, HU]) were similar to those reported by Simpfendorfer et al. [7] (minimal fat renal angiomyolipoma = 37 HU [combined readers]; RCC = 31 HU [combined readers]). Pixel Analysis Radiologist 1 and radiologist 2 had equivalent results (p = across all cohorts and attenuation thresholds) despite a consistently larger ROI size used by radiologist 1 (p < 0.001, Table 1). Table 2 illustrates the number of TABLE 1: Characteristics of Cohorts and Minimal Fat Renal Angiomyolipomas, Clear Cell Renal Cell Carcinomas (RCCs), and Papillary RCCs Minimal Fat Renal Angiomyolipoma Clear Cell RCC Papillary RCC Characteristic Mean Minimum Maximum Mean Minimum Maximum Mean Minimum Maximum Noise (HU) mas kv Age (y) CT date (mo/y) NA 9/02 8/07 NA 8/02 9/06 NA 11/02 7/07 Slice thickness (mm) Mass size (cm) Mean attenuation (HU) Radiologist 1 ROI area (mm 2 ) 0.96 a a a Pixels per ROI Radiologist 2 (mm 2 ) ROI area 0.56 a a a Pixels per ROI Note NA = not applicable, ROI = region of interest. a ROI area was statistically significantly different between radiologist 1 and radiologist 2 within each cohort (p < for all three comparisons). TABLE 2: Number of Masses Within Each Cohort With Histogram Measurements Below Various Attenuation and Pixel-Counting s Attenuation Radiologist 1 > 5% of Pixels Below > 7% of Pixels Below > 10% of Pixels Below > 15% of Pixels Below > 20% of Pixels Below mfaml ccrcc prcc mfaml ccrcc prcc mfaml ccrcc prcc mfaml ccrcc prcc mfaml ccrcc prcc 30 HU HU HU HU HU HU Radiologist 2 5 HU HU HU HU HU HU HU HU Note mfaml = minimal fat renal angiomyolipoma (n = 20), ccrcc = clear cell renal cell carcinoma (n = 22), prcc = papillary renal cell carcinoma (n = 23). AJR:198, February

4 Chaudhry et al. TABLE 3: Combined Radiologist Test Characteristics for the Diagnosis of Minimal Fat Renal Angiomyolipoma at Multiple Attenuation and Pixel-Counting s Performance Value Any % Pixels 5% Pixels 10% Pixels 15% Pixels 20% Pixels 0 HU Sensitivity 0.65 (13/20) 0.40 (8/20) 0.15 (3/20) 0.10 (2/20) 0.05 (1/20) Specificity 0.07 (3/45) 0.51 (23/45) 0.82 (37/45) 0.91 (41/45) 0.96 (43/45) PPV 0.24 (13/55) 0.27 (8/30) 0.27 (3/11) 0.33 (2/6) 0.33 (1/3) NPV 0.30 (3/10) 0.66 (23/35) 0.69 (37/54) 0.69 (41/59) 0.69 (43/62) 5 HU Sensitivity 0.60 (12/20) 0.25 (5/20) 0.10 (2/20) 0.05 (1/20) 0.05 (1/20) Specificity 0.11 (5/45) 0.78 (35/45) 0.91 (41/45) 0.98 (44/45) (45/45) PPV 0.23 (12/52) 0.33 (5/15) 0.33 (2/6) 0.50 (1/2) (1/1) NPV 0.38 (5/13) 0.70 (35/50) 0.69 (41/59) 0.70 (44/63) 0.70 (45/64) 10 HU Sensitivity 0.60 (12/20) 0.15 (3/20) 0.05 (1/20) 0.05 (1/20) 0.05 (1/20) Specificity 0.18 (8/45) 0.84 (38/45) 0.93 (42/45) (45/45) (45/45) PPV 0.24 (12/49) 0.30 (3/10) 0.25 (1/4) (1/1) NA (0/0) NPV 0.50 (8/16) 0.69 (38/55) 0.69 (42/61) 0.70 (45/64) 0.69 (45/65) 15 HU Sensitivity 0.55 (11/20) 0.05 (1/20) 0.00 (0/20) 0.00 (0/20) 0.00 (0/20) Specificity 0.36 (16/45) 0.91 (41/45) 0.98 (44/45) (45/45) (45/45) PPV 0.28 (11/40) 0.20 (1/5) 0.00 (0/1) NA (0/0) NA (0/0) NPV 0.64 (16/25) 0.68 (41/60) 0.69 (44/64) 0.69 (45/65) 0.69 (45/65) 20 HU Sensitivity 0.45 (9/20) 0.05 (1/20) 0.00 (0/20) 0.00 (0/20) 0.00 (0/20) Specificity 0.49 (22/45) 0.98 (44/45) (45/45) (45/45) (45/45) PPV 0.28 (9/32) 0.50 (1/2) NA (0/0) NA (0/0) NA (0/0) NPV 0.67 (22/33) 0.70 (44/63) 0.69 (45/65) 0.69 (45/65) 0.69 (45/65) 25 HU Sensitivity 0.35 (7/20) 0.05 (1/20) 0.00 (0/20) 0.00 (0/20) 0.00 (0/20) Specificity 0.62 (28/45) 0.98 (44/45) (45/45) (45/45) (45/45) PPV 0.29 (7/24) 0.50 (1/2) NA (0/0) NA (0/0) NA (0/0) NPV 0.68 (28/41) 0.70 (44/63) 0.69 (45/65) 0.69 (45/65) 0.69 (45/65) 30 HU Sensitivity 0.25 (5/20) 0.00 (0/20) 0.00 (0/20) 0.00 (0/20) 0.00 (0/20) Specificity 0.76 (34/45) 0.98 (44/45) (45/45) (45/45) (45/45) PPV 0.31 (5/16) 0.00 (0/1) NA (0/0) NA (0/0) NA (0/0) NPV 0.69 (34/49) 0.69 (44/64) 0.69 (45/65) 0.69 (45/65) 0.69 (45/65) Note Data in parentheses yielded performance value. PPV = positive predictive value, NPV = negative predictive value, NA = not applicable. masses in each cohort with pixel counts below various attenuation and pixel-counting thresholds. Minimal fat angiomyolipomas could not be reliably differentiated from RCCs by either radiologist 1 (p = ) or radiologist 2 (p = ) regardless of the chosen attenuation threshold. The combined radiologist sensitivity, specificity, PPV, and NPV of various attenuation and pixel-counting threshold combinations are detailed in Table 3. Independent ROC curves were drawn to illustrate the diagnostic performance of each radiologist at each attenuation threshold (Fig. 1). The paired Student t test showed no significant differences between the two radiologists (Table 4) at any of the attenuation thresholds (i.e., from 30 to 0 HU). Although not statistically significant, there was a trend toward clear cell RCCs being more likely to have low-attenuation pixels than minimal fat renal angiomyolipomas for attenuation thresholds of 0 and 5 HU. Validity of Published Criteria When our data were tested with the methods proposed by Kim et al. [9] and Simpfendorfer et al. [7] for the purpose of distinguishing between 380 AJR:198, February 2012

5 Histogram Analysis of Small Solid Renal Masses Sensitivity Sensitivity HU 5 HU 10 HU 15 HU 20 HU 25 HU 30 HU Specificity HU 5 HU 10 HU 15 HU 20 HU 25 HU 30 HU Specificity Fig. 1 Receiver operating characteristic curves by pixel attenuation threshold. A and B, Radiologist 1 (A) and radiologist 2 (B). A B RCC and minimal fat renal angiomyolipoma, we found discrepant results. We tested multiple pixel-counting (5 20%) and attenuation threshold (0 to 30 HU) combinations and found no reliable method of differentiating minimal fat renal angiomyolipoma from RCC. Criterion 2 from the work by Simpfendorfer et al. would have resulted in radiologist 1 misdiagnosing two RCCs as minimal fat renal angiomyolipomas, and criterion 3 would have resulted in radiologist 1 and radiologist 2 misdiagnosing RCCs as minimal fat renal angiomyolipomas in two cases and one case, respectively. Discussion Small (< 3 cm) solid renal masses lacking macroscopic fat remain a diagnostic dilemma [1]. A minority of these masses are benign (e.g., minimal fat angiomyolipoma, oncocytoma) [1, 7]; however, a noninvasive benign diagnosis is often not possible. Sample CT images from our cohorts that show the similarities in imaging appearances of an minimal fat renal angiomyolipoma, clear cell RCC, and papillary RCC are shown in Figure 2. Various strategies have been proposed to differentiate minimal fat renal angiomyolipoma from RCC on the basis of imaging characteristics. These approaches have included pixel mapping [8], analysis of contrast enhancement patterns [13], and chemical-shift MRI [11]. Despite the initial promise of some of those techniques, none has proven reliable to our knowledge. In several recently published articles, researchers have investigated the usefulness of attenuation measurement histogram analysis with mixed results [7 10]. We revisited this topic by directly evaluating A B C Fig. 2 Sample cases from three cohorts in study. Ovals outline regions of interest. A C, CT scans show minimal fat angiomyolipoma in 53-year-old woman (A), clear cell renal cell carcinoma (RCC) in 48-year-old woman (B), and papillary RCC in 66-yearold woman (C). AJR:198, February

6 Chaudhry et al. the usefulness of several advocated attenuation and pixel-counting thresholds. Kim et al. [9] used attenuation measurement histogram analysis in an attempt to differentiate 34 minimal fat renal angiomyolipomas from 110 size-matched RCCs. One drawback of their study was that only 21 of 34 minimal fat renal angiomyolipomas had histologic confirmation; the remaining cases were classified by a lack of growth over 24 months and 40% signal intensity loss on chemicalshift imaging. They found that minimal fat renal angiomyolipomas in general had lower pixel attenuation values than RCCs using histogram analysis. They also found that minimal fat renal angiomyolipomas could be identified with 100% specificity (20% sensitivity) when the following criteria were used: 6% of pixels or more had a CT number of 10 HU or less. Our results differ from their findings. We tested multiple pixel-counting (5 20%) and attenuation (0 to 30 HU) threshold combinations and found no reliable method of discriminating between minimal fat renal angiomyolipoma and RCC. The findings in our study are concordant with data reported more recently by Catalano et al. [10]. Catalano et al. [10] compared 28 clear cell RCCs with 22 minimal fat renal angiomyolipomas and found that unenhanced CT attenuation measurement histogram analysis was unable to discriminate between the two cohorts. In fact, masses in their study that showed a larger fraction of low-attenuation pixels were actually more likely to be clear cell RCC than minimal fat renal angiomyolipoma. Although TABLE 4: Student t test for Radiologists 1 and 2 our results are not statistically significant, we found a similar trend. In fact, detecting a larger fraction of low-attenuation pixels may be a finding worrisome for malignancy instead of an indicator of benignity. Our study differed from that of Catalano et al. [10] in that their study included only one radiologist who was given specific instructions about how to analyze each renal mass. The radiologist was instructed to draw an oval ROI that encompassed at least three quarters of the lesion at the level of the maximum transverse diameter. This instruction may not reflect how the lesion would be evaluated in the clinical environment. To address this limitation, we allowed the two radiologists in our study to draw ROIs of each mass in the way they would in routine clinical practice. We believe this protocol allowed their training to dictate the method of analysis as opposed to an artificial research-based prescription that may not translate into clinical practice. Simpfendorfer et al. [7] also used unenhanced CT attenuation measurement histogram analysis to try and differentiate a cohort of 18 minimal fat renal angiomyolipomas from a matched cohort of 18 RCCs. They investigated three unique criteria and found that histogram analysis was, in general, inadequate to diagnose the two entities; however, they found two specific criteria that they believed could confidently identify angiomyolipoma with 100% PPV and 100% specificity. These criteria included 20 pixels or more with an attenuation of less than 20 HU (i.e., criterion 2) and 5 pixels or more with an attenuation of less than 30 HU (criterion 3). All minimal fat renal angiomyolipomas correctly classified in this fashion were also found to have greater than 10% fat on histologic examination. We used these attenuation and pixel-counting criteria on our data and found discrepant results. Criterion 3 would have resulted in radiologist 1 and radiologist 2 misdiagnosing RCCs as minimal fat renal angiomyolipomas in two cases and one case, respectively. Criterion 2 would have also resulted in radiologist 1 misdiagnosing two RCCs as minimal fat renal angiomyolipomas. It is possible that our discrepant results are due to the relatively small sample of malignant masses in the study by Simpfendorfer et al. (18 RCCs). We do not recommend the use of either of these criteria for the diagnosis of minimal fat renal angiomyolipoma because of the unacceptable overlap with renal cancers. The limitations of our study include using equipment from a single manufacturer at 140 kvp with a 5-mm slice thickness. It is possible that our results cannot be generalized to images obtained on other manufacturers equipment or at other dose settings. Thinner slices through the target organ could result in increased detection of low-attenuation pixels because of a reduction in partial volume effects. Additionally, six masses (minimal fat renal angiomyolipoma, n = 1; clear cell RCC, n = 3; papillary RCC, n = 2) were scanned using relatively older single-detector CT technology. However, our experience likely reflects that of other institutions in which more than one CT scanner make and mod- Comparison 0 HU 5 HU 10 HU 15 HU 20 HU 25 HU 30 HU Radiologist 1 mfaml (% pixels) ccrcc (% pixels) prcc (% pixels) mfaml vs ccrcc mfaml vs prcc mfaml vs RCC Radiologist 2 mfaml (% pixels) ccrcc (% pixels) prcc (% pixels) mfaml vs ccrcc mfaml vs prcc mfaml vs RCC Note mfaml = minimal fat renal angiomyolipoma (n = 20), ccrcc = clear cell renal cell carcinoma (n = 22), prcc = papillary renal cell carcinoma (n = 23). 382 AJR:198, February 2012

7 Histogram Analysis of Small Solid Renal Masses el is concurrently used in daily clinical practice. All masses were accrued at a single institution and were retrospectively analyzed. These features of our study design may have introduced bias into our cohorts. This possible bias was likely mitigated by matching the minimal fat renal angiomyolipoma cohort to papillary RCC and clear cell RCC cohorts by tumor size, patient sex, and patient age in a fashion similar to that used by Catalano et al. [10] and Simpfendorfer et al. [7]. Last, we did not examine the histologic sections of each resected angiomyolipoma to determine the volume of lipid in each sample as was performed by Simpfendorfer et al. [7]. On the basis of their analysis, they concluded that their criteria could not be applied reliably to angiomyolipomas with less than 10% fat, as measured histologically on biopsy or resection. However, our goal was to analyze a priori methods of correctly distinguishing benign angiomyolipomas from malignant RCCs on the basis of pixel mapping. If we cannot distinguish before resection which minimal fat angiomyolipomas will have a fat fraction of greater than 10%, we cannot rely on criteria that incorporate this feature as a distinguishing characteristic. Further, it is important to note that detection of macroscopic or microscopic fat in a renal mass does not definitively exclude malignancy. Some varieties of clear cell RCC are well known to contain microscopic fat. Cholesterol necrosis, osseous metaplasia, and incorporation of perinephric fat have also been described in rare malignant masses that can simulate angiomyolipoma because of the presence of macroscopic fat. In these rare cases, calcification is often present and indicates malignancy [12, 14 22]. There also is the unusual possibility of an aggressive epithelioid form of angiomyolipoma that has a malignant course. These lesions are typically identified by a pathologic examination and are often found in extrarenal sites with concomitant tumor necrosis [23]. In conclusion, attenuation measurement histogram analysis including the criteria originally offered by Kim et al. [9] ( 6% pixels measuring 10 HU) and those suggested by Simpfendorfer et al. [7] ( 20 pixels measuring 20 HU [criterion 2] or 5 pixels measuring 30 HU [criterion 3]) cannot be used to reliably differentiate minimal fat renal angiomyolipoma from either clear cell RCC or papillary RCC. References 1. Silverman SG, Israel GM, Herts BR, Richie JP. Management of the incidental renal mass. Radiology 2008; 249: Hajdu SI, Foote FW Jr. Angiomyolipoma of the kidney: report of 27 cases and review of the literature. J Urol 1969; 102: Said JW, Thomas G, Zisman A. Kidney pathology: current classification of renal cell carcinoma. Curr Urol Rep 2002; 3: Hafron J, Fogarty JD, Hoenig DM, Li M, Berkenblit R, Ghavamian R. Imaging characteristics of minimal fat renal angiomyolipoma with histologic correlations. Urology 2005; 66: Jinzaki M, Tanimoto A, Narimatsu Y, et al. Angiomyolipoma: imaging findings in lesions with minimal fat. Radiology 1997; 205: Frank I, Blute ML, Cheville JC, et al. Solid renal tumors: an analysis of pathological features related to tumor size. J Urol 2003; 170: Simpfendorfer C, Herts BR, Motta-Ramirez GA, et al. Angiomyolipoma with minimal fat on MDCT: can counts of negative-attenuation pixels aid diagnosis? AJR 2009; 192: Simpson E, Patel U. Diagnosis of angiomyolipoma using computed tomography-region of interest < or = 10 HU or 4 adjacent pixels < or = 10 HU are recommended as the diagnostic thresholds. Clin Radiol 2006; 61: Kim JY, Kim JK, Kim N, Cho KS. CT histogram analysis: differentiation of angiomyolipoma without visible fat from renal cell carcinoma at CT imaging. Radiology 2008; 246: Catalano OA, Samir AE, Sahani DV, Hahn PF. Pixel distribution analysis: can it be used to distinguish clear cell carcinomas from angiomyolipomas with minimal fat? Radiology 2008; 247: Kim JK, Kim SH, Jang YJ, et al. Renal angiomyolipoma with minimal fat: differentiation from other neoplasms at double-echo chemical shift FLASH MR imaging. Radiology 2006; 239: Outwater EK, Bhatia M, Siegelman ES, Burke MA, Mitchell DG. Lipid in renal clear cell carcinoma: detection on opposed phase gradient-echo MR images. Radiology 1997; 205: Kim JK, Park SY, Shon JH, Cho KS. Angiomyolipoma with minimal fat: differentiation from renal cell carcinoma at biphasic helical CT. Radiology 2004; 230: Hélénon O, Chrétien Y, Paraf F, Melki P, Denys A, Moreau JF. Renal cell carcinoma containing fat: demonstration with CT. Radiology 1993; 188: Hélénon O, Merran S, Paraf F, et al. Unusual fatcontaining tumors of the kidney: a diagnostic dilemma. RadioGraphics 1997; 17: Hammadeh MY, Thomas K, Philp T, Singh M. Renal cell carcinoma containing fat mimicking angiomyolipoma: demonstration with CT scan and histopathology. Eur Radiol 1998; 8: Cribbs RK, Ishaq M, Arnold M, O Brien J, Lamb J, Frankel WL. Renal cell carcinoma with massive osseous metaplasia and bone marrow elements. Ann Diagn Pathol 1999; 3: D Angelo PC, Gash JR, Horn AW, Klein FA. Fat in renal cell carcinoma that lacks associated calcifications. AJR 2002; 178: Schuster TG, Ferguson MR, Baker DE, Schaldenbrand JD, Solomon MH. Papillary renal cell carcinoma containing fat without calcification mimicking angiomyolipoma on CT. AJR 2004; 183: Kefeli M, Yildiz L, Aydin O, Kandemir B, Faik Yilmaz A. Chromophobe renal cell carcinoma with osseous metaplasia containing fatty bone marrow element: a case report. Pathol Res Pract 2007; 203: Hayn MH, Cannon GM, Bastacky S, et al. Renal cell carcinoma containing fat without associated calcifications: two case reports and review of literature. Urology 2009; 73:e5 e7 22. Aron M, Aydin H, Sercia L, Magi-Galluzzi C, Zhou M. Renal cell carcinomas with intratumoral fat and concomitant angiomyolipoma: potential pitfalls in staging and diagnosis. Am J Clin Pathol 2010; 134: Lai HY, Chen CK, Lee YH, Tsai PP, Chen JH, Shen WC. Multicentric aggressive angiomyolipomas: a rare form of PEComas. AJR 2006; 186: AJR:198, February